Synthetic Media in Multilingual MOOCs: Deepfake Tutors, Pedagogical Effects, and Ethical-Policy Challenges

In recent years, synthetic media from deepfake videos have emerged as a new interesting technology, whether that refers to cloned voices, multilingual translation models, or more recent applications of avatar tutors into higher education. As such, these technologies are rapidly becoming part of the multilingual distance learning model and, more recently, MOOCs worldwide. This article is a scoping review that focuses on recent international literature published between 2020 and 2025 to explore the usage of deepfake and synthetic media tools and methods in multilingual MOOC content and assess the influence of these technologies on social presence and participation. Similarly, we focus on ethical and political issues that are closely connected with the adaptation of these technologies, and upon analysing educational technology and policy documents, such as UNESCO’s Guidelines and the EU AI Act, we pinpoint that the use of synthetic avatars and AI-generated videos can diminish production costs and assist multilingual learning. Evidently, concerns arise regarding authenticity, privacy, and the shifting nature of the teacher-learner relationship that are thoroughly discussed. As a result, the technical merit of this paper is the proposal of a policy framework that, in an effort to address these issues, focuses on transparency, responsible governance, and AI literacy. The goal is not to replace human instruction but to integrate synthetic media in ways that strengthen pedagogical design, safeguard rights, and ensure that multilingual MOOCs become more interesting and inclusive rather than more automated robotic processes and unequal

💡 Research Summary

This paper presents a comprehensive scoping review of the emerging use of synthetic media—including deepfake videos, voice‑cloning, avatar tutors, and speech‑to‑speech translation systems—in multilingual Massive Open Online Courses (MOOCs) published between 2020 and 2025. The authors adopt the Joanna Briggs Institute (JBI) methodology and PRISMA‑ScR reporting standards, defining a PCC (Population‑Concept‑Context) framework that targets students, instructors, and MOOC designers (Population); deepfake and synthetic‑media technologies such as AI‑generated avatars, cloned voices, text‑to‑video, and multimodal translation (Concept); and multilingual or international MOOCs (Context). A systematic search across IEEE Xplore, ACM Digital Library, Scopus, Web of Science, SpringerLink, Taylor & Francis, as well as grey literature from UNESCO and EU policy documents, yielded 112 relevant studies and reports for analysis.

The review is organized around three research questions: (1) What forms of deepfake and synthetic media are being employed in multilingual MOOCs? (2) What are the pedagogical implications of these technologies? (3) What ethical, cultural, and political challenges arise, and how can policy address them?

Technological Landscape (RQ1). The authors categorize synthetic media into four main families: (a) AI‑driven avatar tutors that replicate a human instructor’s facial expressions, gestures, and vocal prosody; (b) voice‑cloning systems that generate multilingual speech from a single recorded voice; (c) multimodal translation pipelines such as SEAMLESSM4T that provide real‑time speech‑to‑speech conversion in up to 100 languages; and (d) text‑to‑video generators that produce visual explanations from textual scripts. These tools rely on large language models (LLMs), generative adversarial networks (GANs), diffusion models, and transformer‑based speech synthesis, enabling a single instructor to produce cost‑effective, culturally adapted content at scale.

Pedagogical Effects (RQ2). Drawing on the Community of Inquiry (CoI) framework, the paper evaluates impacts on social presence, cognitive presence, and teaching presence. Empirical findings across the reviewed literature suggest that synthetic avatars can increase learner engagement and personalization, especially for non‑English speakers, by offering immediate feedback and language‑matched delivery. However, the absence of authentic non‑verbal cues (eye contact, facial micro‑expressions, natural prosody) often reduces perceived social presence and emotional connection. One large‑scale experiment cited (n = 447) reported comparable learning outcomes between AI‑generated and human‑taught videos, yet a statistically significant 12 % drop in social presence scores for the synthetic condition. The authors argue that transparent disclosure of synthetic agents mitigates trust erosion, while blended designs—human instructor plus AI‑augmented segments—preserve the benefits of both modalities.

Ethical and Policy Dimensions (RQ3). The review foregrounds four interrelated concerns: (i) transparency (mandatory labeling of AI‑generated content), (ii) privacy (informed consent for the use of personal voice or facial data), (iii) data sovereignty (clear retention and deletion policies), and (iv) risk management (classification of educational AI as “high‑risk” under the EU AI Act 2024/1689). UNESCO’s 2023 Guidance for Generative AI in Education is examined alongside the EU AI Act, both of which stress human‑centred design, safety, accountability, and AI literacy. The authors propose a policy framework built on three pillars: (1) Transparent Disclosure – every synthetic video, avatar, or voice must carry a digital watermark and an accessible “About this AI” page; (2) Informed Consent & Data Governance – institutions must obtain explicit consent from instructors and learners before collecting biometric data, and must implement secure storage, audit trails, and right‑to‑erase mechanisms; (3) AI Literacy Programs – mandatory curricula for both learners and educators that cover the technical basics of generative AI, ethical considerations, and critical evaluation skills.

To operationalize these pillars, the paper recommends establishing a Synthetic Media Governance Committee within each MOOC provider or university. This body would conduct pre‑deployment ethical impact assessments, monitor compliance with labeling and consent requirements, and oversee periodic audits aligned with UNESCO and EU standards. The committee would also serve as a liaison with regulators, ensuring that any updates to the AI Act or UNESCO guidelines are promptly integrated into institutional policies.

Methodological Rigor. The authors detail their adherence to JBI’s six‑step scoping review process (question formulation, searching, selection, charting, summarizing, reporting) and PRISMA‑ScR checklists, ensuring reproducibility. Inclusion criteria required peer‑reviewed articles, conference papers, or official reports that explicitly addressed synthetic media in the context of multilingual MOOCs. Exclusion criteria filtered out works focusing solely on security threats, political misinformation, or monolingual settings. Data extraction captured technology type, study design, sample size, measured outcomes (learning performance, engagement, presence), and policy recommendations.

Conclusions and Contributions. The review concludes that synthetic media hold substantial promise for reducing production costs, enhancing linguistic equity, and personalizing learning at scale. Nevertheless, without robust governance, these tools risk undermining social presence, authenticity, and privacy, potentially widening rather than narrowing educational inequities. The authors’ fourfold contribution includes (1) integrating pedagogical and ethical analyses, (2) centering multilingual MOOCs and linguistic justice, (3) mapping international regulatory frameworks (UNESCO, EU AI Act), and (4) advancing the concept of “synthetic pedagogy” that positions AI as an augmentative, not replacement, element. By adopting the proposed transparency‑governance‑literacy framework, institutions can harness synthetic media to create more inclusive, engaging, and ethically sound multilingual MOOCs.